This review currently includes 19 randomised studies with a total of 2443 participants. Most trials were of short duration. Data from 4 trials with 514 participants with predominantly negative symptoms suggest that low-dose (up to 300mg/day) amisulpride was a more acceptable treatment than placebo (n=514, RR 0.6 CI 0.5 to 0.8, NNT 3 CI 3 to 7), the improvement of the participants' global state (n=242, RR 0.6 CI 0.5 to 0.8, NNT 3 CI 2 to 6) and the treatment of negative symptoms (n=177, WMD -10.1 CI -16.6 to -3.5). Amisulpride was shown to be more likely to cause extrapyramidal symptoms than placebo in two studies (n=269, RR 2.2 CI 1.2 to 4.2), but this result did not hold calculating the risk reduction so that an NNT-statistic could not be indicated.
Compared to typical antipsychotics, the pooled results of a total of fourteen trials suggest that amisulpride was more effective in improving global state (n=651, RR 0.7 CI 0.5 to 0.9, NNT 6 CI 4 to 11), the general mental state (n=695, WMD -4.2 CI -6.5 to -1.9) and the negative symptoms of schizophrenia (n=506, WMD -2.8 CI -4.3 to -1.3). Regarding positive symptoms, amisulpride was as effective as typical antipsychotics. Amisulpride was less prone to cause at least one general adverse event (n=751, RR 0.9 CI 0.8 to 0.97, NNH 9 CI 6 to 18), one extrapyramidal symptom (n=771, RR 0.7 CI 0.6 to 0.9, NNH 5 CI 4 to 9) or to require the use of antiparkinson medication (n=851, RR 0.6 CI 0.5 to 0.8, NNH 4 CI 3 to 6). No clear differences in other adverse events compared to typical drugs were found. Amisulpride also seemed to be more acceptable than conventional drugs as measured by the outcome 'leaving the studies early' (n=1512, RR 0.8 CI 0.7 to 0.9, NNT 16 CI 9 to 69) than conventional drugs, but this result might have been overestimated due to a publication bias which could not be excluded with certainty.
A single trial compared amisulpride to another 'atypical' antipsychotic, risperidone. With the exception of agitation, which was more frequent in the amisulpride group (n=228, RR 3.4 CI 1.2 to 10.1, NNH 11 CI 6 to 50) no significant differences were recorded on efficacy or acceptability.